4.2 The Andersen report - and the costs of capital

However this is by no means the end of the story. As we have seen, the Andersen report assumed that the PFI cost of capital would be only one to three percentage points above the public sector borrowing rate (Andersen et al 2000, 5). But the actual gap is in fact much greater.

Firstly let us look more closely at the PFI cost of capital. PFI projects are financed by a mix of capital types, namely senior debt, subordinated loan (or junior) capital and equity capital. It is the senior debt which will be at the lowest rate of interest and as the Andersen report put it (in a footnote on their page 10); "…higher returns will be demanded for junior debt and equity finance" (Andersen et al 2000, fn 2 on page 10). Indeed this seems to be the case, as pointed out by the Audit Scotland/Accounts Commission in 2002. They quoted the following costs of capital for six Scottish PFI school projects;

• senior debt capital 6-7% pa

• subordinated loan capital 10-16% pa

• equity capital 15-29% pa

• weighted average 7-13% pa
(Audit Scotland/Accounts Commission, 2002, 58)

The weighted average is closer to the rate on senior debt capital than that on equity capital because senior debt capital usually accounts for around 90% of funding (Audit Scotland/Accounts Commission, 2002, 57). At the same time it is important to emphasise that the return on equity is unlimited - it will depend on the terms of the contract. But if we take the mid-point of the above weighted average, then we have a rate of 10% pa and since this seems to be in nominal terms, this agrees with the assumption of the PFI cost of capital adopted by the Andersen report.

However it is somewhat below the average real cost of capital of thirteen PFI companies (SPVs or Special Purpose Vehicles) in the recent study by Pam Edwards et al. The latter study showed an average rate of profit of well over 100% pa on the equity capital (shareholders' funds) in each of the years between 2000 and 2002 inclusive but because of the very high (more than 95%) gearing ratios (debt to total capital) of the SPVs in the health sector, the average cost of capital was 7% in 2000 and 10% in each of 2001 and 2002 (Edwards et al 2004, 167). This would seem from their table to be in real terms and yields an average of about 9% pa (in real terms)14 or about 5 percentage points above the Treasury stock, as Edwards et al point out on pages 11 and 174 of their report.

In a more recent study of 12 hospital schemes, the average rate of interest paid on debt between 2002 and 2005 inclusive was 7.8% pa (Shaoul et al 2007, table 5). In the same study the rate of profit on equity capital in 2005 averaged 58% pa. Even assuming a very high debt-to-total-PFI ratio of 95%, the average real cost of capital is at least 10% pa. Similarly the annual real cost of capital at the Norfolk and Norwich University Hospital (NNUH) has been almost exactly 10% per annum. By contrast, Andersen had assumed a nominal rate of 10% pa and a real rate of 7.5% pa. Therefore the latter seems to be about 2.5% too low and as a result their assumption was heavily biased in favour of the PFI.

However Andersen made another mistaken assumption, this time about the public sector borrowing rate. The real rate assumed by Andersen was 6% pa which is too high. For a start, as pointed out earlier, this includes an allowance for risk and the correct rate, excluding risk, should be 4.5% pa. Even this latter figure is a bit too high inasmuch as it compares with a real average yield on government bonds which is slightly lower. The Andersen report was produced in the year 2000 so it is reasonable to look at the annual yield over the 20-year period from 1981 to the year 2000 inclusive. This is done in table 4.2 below.

The annual average yield (in real terms) is shown to be 4.3% pa, so that the real interest rates that should be plugged into the model are 10% pa for the PFI cost of capital and 4.3% pa for the Public Sector Comparator.

These rates are used in table 4.3 to compare the costs of capital of the PFI version with the PSC using a discount rate of 6% pa.

The more realistic assumptions adopted in table 4.3 give much more of an anti-PFI picture than given in Table 4.1. The capital cost of using the private sector for financing the hospital is now seen to be £140.8 mn compared to £84.4 mn using public finance. Thus the PFI is £56.4 mn more expensive. In other words, for the same base construction cost, the PFI version would cost 67% more than the publicly-financed alternative. To give value for money, the construction cost of the PFI version has to be at least 40% lower than that of the publicly-financed version. In other words, if we look at the Andersen model of a construction cost of £100 mn if publicly financed, we can see that in order to compensate for its higher interest cost, the private sector's construction cost would have to be £60 mn.

It is clear that the relative real interest rates (7.5% pa for the private sector and 6% for the public sector) assumed by the Andersen report were highly misleading. Realistic real interest rates (10% pa for the private sector and 4.3% for the public sector) mean that the private sector's base construction cost has to be 40% lower than that of the public sector compared to 13% if we use the interest rates assumed by Andersen.

Thus the figures given in the Andersen report provide a very distorted picture. George Monbiot in chapter 2 of The Captive State suggested that the Andersen "report was an honest attempt to answer the questions the Treasury posed" (Monbiot, 2000, 81). It is questionable whether it was honest, but it was certainly not accurate. Because of its wrong assumptions of interest rates, it painted far too rosy a picture of the PFI.

14 . Much higher average rates of return on equity and debt capital are quoted elsewhere in connection with the PFI. For example Serco has stated that "after-tax returns anticipated on equity and lending will vary according to the perceived risk profile, but are currently 15-20% on capital invested" (SERCO 2004, 46).